Wavelet-Based Simulation Model Validation of Functional Data
Abstract
As computer hardware technology continues to advance, so does the scientific community's capability to develop high resolution computer models able to simulate complex systems and processes. This advancement has led to many challenges associated with verification and validation (V and V). These challenges include adapting methods to high dimensional functional data, maintaining the necessary objectivity, and accounting for noisy data. Department of Defense (DoD) simulation models require validation techniques that are able to overcome these challenges before the models can be relied upon. Model validation substantiates that the model chosen sufficiently represents the system and that it produces results consistent with real-world data within the range of model applicability. In this research, new statistical techniques will be proposed that improve upon existing simulation validation techniques. These techniques incorporate the use of wavelets to decompose the time-series data into the time-frequency spectrum allowing for objective and comprehensive assessment of the model. In addition, these techniques offer an improved method of analysis for noisy, high-dimensional data. These techniques are applied to assess the validity of simulation models, which will help ensure the accurate representation of the system they are meant to simulate.